Stanford's New CS 312 "Deep Learning Alchemy" Will Release All Materials Publicly
Stanford NLP will launch CS 312 "Deep Learning Alchemy" in fall 2026, taught by Tatsunori Hashimoto and Suhas Kotha, with Kaiyue Wen as course assistant. All lectures and materials will be publicly released, and grading is unusually radical: homework counts for 0% while 85% of the grade comes from a 40-minute exam. The course is already full with a waitlist open, starting in about 10 days.
Confirmed
- Instructors are Tatsunori Hashimoto and Suhas Kotha, with Kaiyue Wen as CA (corroborated by multiple posts)
- All recordings and course materials will be made public, following the cs336 open style
- Homework is 0% of the grade; 85% comes from a 40-minute exam where students must predict outcomes of experimental setups they have never run
- The course is full with a waitlist open
- The design philosophy is "prediction is understanding" and "alchemy can only be learned hands-on," designed by Hashimoto of Percy Liang's team
Why it matters
- Making "predicting experiment outcomes" the core measure of understanding is a bold restructuring of homework-centric deep learning teaching
- Full public release means anyone can self-study, offering a window into the Stanford NLP group's pedagogical approach
2026-09-13 ~ 2026-09-14 · 5 related posts
Primary sources
- [source] Stanford's new CS 312 grades 85% on predicting experiment outcomes, 0% on homework — stanfordnlp · 2026-09-13
- Stanford's CS 312 'Deep Learning Alchemy' by Hashimoto & Kotha to release all materials publicly — anshulkundaje · 2026-09-13
- Stanford's new CS 312 'Deep Learning Alchemy' course makes all materials public — stanfordnlp · 2026-09-14
2 near-duplicate retellings: PMinervini · JeremyNguyenPhD